key: cord-0253289-s9knq3q4
authors: Duchene, S.; Featherstone, L.; de Blasio, B. F.; Holmes, E. C.; Bohlin, J.; Pettersson, J. H.- O.
title: Assessment of COVID-19 intervention strategies in the Nordic countries using genomic epidemiology
date: 2021-09-08
journal: nan
DOI: 10.1101/2021.09.04.21263123
sha: 1ac0302c471c40c854cafab08b10c35900e3f2e3
doc_id: 253289
cord_uid: s9knq3q4

The Nordic countries, defined here as Norway, Sweden, Denmark, Finland and Iceland, are known for their comparable demographics and political systems. Since these countries implemented different COVID-19 intervention strategies, they provide a natural laboratory for examining how COVID-19 policies and mitigation strategies affected the propagation, evolution and spread of the SARS-CoV-2 virus. We explored how the duration, the size and number of transmission clusters, defined as country-specific monophyletic groups in a SARS-CoV-2 phylogenetic tree, differed between the Nordic countries. We found that Sweden had the largest number of COVID-19 transmission clusters followed by Denmark, Norway, Finland and Iceland. Moreover, Sweden and Denmark had the largest, and most enduring, transmission clusters followed by Norway, Finland and Iceland. In addition, there was a significant positive association between transmission cluster size and duration, suggesting that the size of transmission clusters could be reduced by rapid and effective contact tracing. Thus, these data indicate that to reduce the general burden of COVID-19 there should be a focus on limiting dense gatherings and their subsequent contacts to keep the number, size and duration of transmission clusters to a minimum. Our results further suggest that although geographical connectivity, population density and openness influence the spread and the size of SARS-CoV-2 transmission clusters, country-specific intervention strategies had the largest single impact.

interventions that included a widespread population "lock-down". Iceland, a small isolated island 71 population, focused on contact tracing and testing (1). Because of their homogeneity, the Nordic 72 countries therefore provide a natural laboratory to compare COVID-19 intervention strategies (2, 3). 73

We recently analysed transmission clusters, defined as country-specific monophyletic groups 74 derived from a phylogenetic analysis of the causative Severe Acute Respiratory Syndrome 75

Coronavirus 2 (SARS-CoV-2), in the Nordic countries using a genomic data set covering the 76 Nordic region (1). Here, we use the same transmission clusters to analyse the epidemiological 77 implications of the different intervention strategies employed by the Nordic health authorities. 78

Briefly, the data was generated in (1) by (i) downloading and aligning all available Nordic SARS-80

CoV-2 genomes from GISAID (www.gisaid.org; with 67,918 Nordic genomes) and the NextStrain 81 global build as of 22nd March 2021 (3,437 global genomes). The data set was then subsampled ten 82 times (referred to here as "replicates") according to prevalence per country, resulting in alignments 83 including between 15,297 and 15,616 SARS-CoV-2 genome sequences. For each alignment, a 84 phylogenetic tree was computed with IQ-TREE v2.0.6 (4) scaled to time using LSD v.03 (5) under 85 a strict molecular clock at a fixed rate of 1×10 -3 nucleotide substitutions/site/year, and the GTR+Γ 86 substitution model. Transmission clusters were defined as the monophyletic clustering of two or 87 more sequences from the same country. The duration of a transmission cluster was defined as the 88 time between the last and the first sample. The data set was converted to a weekly time-series 89 format, calculated from the time to most recent common ancestor (TMRCA) from each 90 transmission cluster. Analyses were carried out using generalised additive mixed-effects models 91 (GAMM) employing restricted maximum likelihood (REML) (6). We used GAMM for the 92 regression models with country as the explanatory variable and the following outcomes: number of 93 transmission clusters, transmission cluster size (i.e. the number of samples in each transmission 94 cluster) and transmission cluster duration. An additional GAMM model was included to assess the 95 association between duration and transmission cluster size. This model included an interaction term 96 between country and number of samples as the explanatory variable with duration as the outcome. 97

All models, except that having the number of transmission clusters as an outcome, were also 98 adjusted for the number of transmission clusters and TMRCA with respect to country using splines. 99

Replicate number (1-10) by country (Norway, Sweden, Finland, Denmark and Iceland) was 100 included as a hierarchical random effect in all models. Goodness-of-fit was determined using 101

Akaike Information Criterion (AIC), R 2 and model residuals conformance to normality. Regression 102 analyses were performed using the GAMM4 package (6), and figures were created using the 103 ggplot2 package (7). A description of the statistical estimation procedures and results are provided is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 8, 2021. ;

We divided the data set into the number of transmission clusters per week. First, we compared the 107 number of transmission clusters between the different countries (see Supplementary Figure S1 and 108 the Methods section for more details regarding the regression models). This revealed that Sweden 109 had significantly more transmissions clusters (p<0.001) than Denmark, who in turn had 110 significantly more than Norway (p<0.001). The number of transmission clusters in Finland was 111 comparable to Norway (p=0.878). The fewest number of transmission clusters were found in 112

Iceland (p<0.012), but the smoothing spline did not fit the data properly (p=0.906) due to the low 113 number of cases. 114

We next compared transmission cluster size (i.e. number of samples per cluster) between the 115 different Nordic countries. These analyses suggest that transmission cluster size did not differ 116 significantly between Sweden and Denmark (p=0.895). However, transmission clusters were 117 significantly smaller (p<0.001) in Norway, followed by Finland (p<0.001), while Iceland's model 118

was not found to be significant (p=0.898). Figure We also compared transmission cluster duration between the Nordic countries. This did not 124 significantly differ between Sweden and Denmark (p=0.269) who, in turn, had significantly more 125 enduring transmission clusters than the other Nordic countries (p<0.002). There was no significant 126 difference between Norway and Finland (p=0.503), while Iceland's model was again not significant 127 (p=0.887) due to the low number of cases. In Figure 2 we see that the duration of the transmission 128 clusters drops when the winter months approach the end of 2020. A significant positive association 129 (p<0.001) between transmission cluster size and duration was established suggesting that the 130 duration of a transmission cluster correlates with its size. 131

We examined how the number of transmission clusters and their size and duration, differed among 133 the Nordic countries. The presence of many samples within a transmission cluster indicate that one 134 common source is responsible for many subsequent cases, while a low number points to fewer 135 secondary infections. Accordingly, Sweden had the largest number of transmission clusters and, 136 together with Denmark, also the most enduring. Moreover, our results also point to an increase of 137 COVID-19 transmission clusters during winter, which could be due to the cold winter climate 138 leading to increased indoor-based activities and thus increased contact frequencies for transmission 139 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 8, 2021. ; https://doi.org/10.1101/2021.09.04.21263123 doi: medRxiv preprint (8). Alternatively, the increased spread of SARS-CoV-2 during the winter months could result from 140 more relaxed COVID-19 intervention measures during the autumn (1). Notably, transmission 141 cluster size decreased towards the end of 2020 in concordance with an increase in the governmental 142 stringency index for the Nordic countries (1). Hence, while climate and change of season influence 143 behaviour and contact frequency, our analyses suggest that the adopted intervention measures also 144 effectively reduced virus transmission. 145

The largest transmission clusters were found in Sweden and Denmark, while the size of the 146 transmission clusters in Norway and Finland were significantly smaller and more similar. Iceland 147 differed in general from the other Nordic countries with substantially fewer COVID-19 cases, 148 which can most likely be ascribed to the country's small population and isolated location. 149

Interestingly, although Sweden had more transmission clusters (Supplementary Figure S1 ), their 150 duration and size was not significantly different from the clusters obtained from Denmark. 151

Although we have tried to adjust for bias in sequencing intensity by using sub-sampled SARS-CoV-152 2 sequence data from each country in the regression models (1), some bias may still persist due to 153 country-specific strategies for selecting cases for sequencing. In particular, Swedish transmission 154 clusters could be underestimated in terms of both size and duration. Alternatively, the more 155 stringent intervention policies implemented in Denmark, as compared to Sweden, could have 156 reduced the total number of infections but allowed the transmission clusters to endure. Indeed, we 157 found a significant association between transmission cluster size and duration. Finland and Iceland 158 had the most short-lived transmission clusters, with Iceland's significantly shorter than Finland's. 159

During the second half of 2020, Sweden adopted stricter COVID-19 intervention policies, more on 160 par with its neighbouring countries Norway and Finland, but the number of cases remained high (1). 161

This may point to some inertia regarding COVID-19 policies, such that the attitudes in the 162 population and consequences from early COVID-19 intervention strategies persisted for some time 163 after the new measures were introduced. As Norway and Finland are similar to Sweden in terms of 164 demographics, location, climate and governance, our results demonstrate the effects of the variable 165 COVID-19 control measures adopted. Norway, Finland and Denmark experienced a similar burden, 166 while significantly more cases and transmission clusters were noted for Denmark. The increased 167 number of cases and transmission clusters observed in Denmark, compared to Norway and Finland, 168 could in part be due to far more intensive SARS-CoV-2 sequencing, although it may also be a 169 consequence of Denmark's higher population density and placement on the European continent. We 170 observed a similar pattern in our previous study (1) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 8, 2021. ;  significantly shorter compared to Denmark and Sweden. The Norwegian and Finnish strategies 174 have focused on municipality-based, rigorous contact tracing, isolation and quarantining with the 175 aim to clamp down transmission. 176

Our results suggest that reducing transmission cluster duration through effective contact tracing 177 could also reduce transmission cluster size. The shorter duration of transmission clusters in Norway 178

and Finland may reflect the effectiveness of the strategies deployed in those countries. Importantly, 179 a growing number of transmission clusters would likely imply an increase in the genetic diversity of 180 SARS-CoV-2, while the presence of transmission clusters of extended length better enable the virus 181 to evolve. 182

The SARS-CoV-2 Alpha variant of concern (Pangolin lineage B.1.1.7) was first discovered in the 183 United Kingdom and was associated with an increase in infectivity (9). Interestingly, we did not see Denmark may be more challenging as it is a relatively small country, located on the European 199 continent, with a comparatively high population density. In contrast, the pattern in Iceland may 200 reflect the fact that it is an island with a substantially smaller population located far away from 201 other countries. Our findings from Iceland resembled those of other small island nations (10). 202

Using a genomic epidemiology-based approach we found that transmission cluster size and duration 204 differed markedly among the Nordic countries. Considering the homogeneous populations and the 205 similar political systems in these countries, our findings clearly demonstrate that COVID-19 206 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 8, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 intervention policies can have profound influence on how epidemics evolve at the scale of 207 individual countries. In particular, as Norway, Sweden and Finland are neighbouring countries with 208 similar demographics, political systems and climate, the data presented here highlight the influence 209 of intervention strategies in controlling the spread of COVID-19. Interestingly, Denmark 210 implemented more similar COVID-19 intervention strategies to Norway and Finland, but did not 211 reduce the spread of COVID-19 to same extent as other Nordic countries which may reflect its 212 location on the European continent and relatively high population density. Hence, our results 213

suggest that disease intervention strategies should be adapted to the specific geographic and 214 demographic factors of each country to more effectively reduce the transmission and evolution of 215 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 8, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 8, 2021. ; https://doi.org/10.1101/2021.09.04.21263123 doi: medRxiv preprint

The 219 impact of early public health interventions on SARS-CoV-2 transmission and evolution

Elimination could be the optimal response strategy for 222

covid-19 and other emerging pandemic diseases

SARS-CoV-2 elimination, not mitigation, 224 creates best outcomes for health, the economy, and civil liberties

IQ-TREE 2: New Models and Efficient 227

Methods for Phylogenetic Inference in the Genomic Era

Fast Dating Using Least-Squares Criteria and 230

Generalized additive models: an introduction with R

Package 'ggplot2'. Create Elegant Data 233